Google is rolling out important updates to its generative AI chatbot, Bard, making it more adaptable and integrated with Google’s ecosystem of apps and services. The latest version of Bard introduces extensions, the capacity to verify responses, and alliance attributes. These enhancements are designed to provide users with a more extensive and reliable AI chat experience.
Extensions: Integrating with Google Services One of the standout features of the updated Bard chatbot is its combination with Google’s suite of apps and services, which includes Gmail, Docs, Drive, Maps, YouTube, and Google Flights and Hotels. However, this combination comes with a strong focus on user trust and privacy. Users must explicitly opt-in to permit Bard to interact with these services, and they can rescind that permission at any time. Importantly, personal data from these services is not used for reinforcement learning, ensuring user privacy and data security. For example, if a user opts in to use Gmail with Bard, the chatbot can find certain information in the user’s inbox as requested but does not gain entry to the entire inbox. Google emphasizes that no human reviewers will have entry to the email content Bard interacts with, maintaining a high level of privacy and trust.
Practical Use Cases of Extensions The extensions feature opens up various practical use cases. For instance, Bard can study and outline important emails, making it easier for users to catch up on critical information. However, it’s essential to note that Bard’s capabilities are dependent on the specificity of the email content. While it can retrieve information from emails, it may not recognize more nuanced details without specific keywords. Extensions can also work jointly seamlessly. In a collaborative trip preparation scenario, Bard can extract relevant dates from email conversations, look up present flight and hotel information, give directions using Google Maps, and even suggest YouTube videos showcasing activities at the destination—all within a single conversation.
Verifying Bard’s Responses Google is taking steps to increase the dependability of Bard’s responses. Users can now double-check Bard’s answers by tapping the “G” icon, which triggers a validation process against Google Search. Each sentence made by Bard is approved to determine if there is web content that supports the response. Users can click on highlighted phrases to access additional information through Google Search. Importantly, Bard’s uncertainty is transparently communicated. If the AI is unsure about a specific part of the response, that portion will be highlighted in orange. This feature helps users identify when Bard may be “hallucinating,” providing responses based on incomplete or incorrect information. It also helps the AI learn from user feedback, improving its accuracy over time.
The latest update allows users to collaborate with others through Bard. When someone shares a Bard chat via a public link, recipients can continue the conversation and ask additional questions on the same topic. This collaborative feature provides a platform for discussions and idea sharing, making Bard a valuable tool for group interactions. Google is expanding Bard’s language support to over 40 new languages, making its features, including image uploads with Google Lens, search images in responses, and response modification, accessible to a more extensive global audience.